• by Laurie Sullivan  @lauriesullivan, February 14, 2025

Dream Screen, an experimental AI feature from YouTube that creates backgrounds for videos from text prompts, is getting a major upgrade with the integration of Google DeepMind's newest video generation model, Veo 2. The company also will launch a feature that generates standalone video clips that creators can add to any of Shorts.

The feature, which the company previewed last year at Made on YouTube, enables creators to build specific scenes by using a text prompt to generate a video clip that fits into the narrative.

The tool also generates videos faster because with Veo 2, the tool understands real-world physics and human movement better, making the output more detailed and realistic. It also allows the creator to use a particular style, lens, or cinematic effect.

The technology uses SynthID watermarks with clear labels to indicate that these creations were generated with artificial intelligence (AI).

These features launched in the U.S., Canada, Australia, and New Zealand, with plans to expand to more of you in the future.

Creators must take responsibility to ensure that published content, including content created using AI tools, follows YouTube’s rules.

YouTube has incorporated safeguards into its AI tools, which aim to help creators follow Community Guidelines and avoid generating inappropriate content.

YouTube may block prompts that violate its policies, or those that touch on sensitive topics. For example, the AI tools are designed to help creators avoid generating photorealistic images of identifiable people, which could be misused for creating deep fakes or other harmful content.

To create video backgrounds for Shorts, users can open the Shorts camera and then select Green Screen and the Dream Screen. From there, the user types in a prompt, selects an image, and then creates the video background.

To create a clip and add it to a Short, users can open the Shorts camera, tap Add to open the media picker, then tap Create at the top. After inputting the prompt, the user selects the image, then taps Create video, and chooses the desired length.